Chapter 155: The Compatibility Layer and the White Paper

The clock in the lower-right corner of the screen flipped to 02:14. The pool of light from the desk lamp had shrunk to the middle of the tabletop, illuminating a scatter of draft pages and half a cup of boiled water gone cold. Lin Chen set down his red pen and rubbed the stiff joints of his fingers. The swelling in his left ankle had already spread up into his calf; the skin was stretched taut and shiny, and when he pressed it, it did not spring back.

He scooted his chair back half a foot and propped his left leg on a stack of folded old quilts. The throbbing caused by poor circulation jabbed into the seams of his bones like fine needles. He closed his eyes and took three deep breaths, waiting for the spasm in the muscle to pass. He could not keep forcing himself through the night. His body was his only real capital; overdraw it once, and recovery would take three extra days. He had set a rule for himself: forty-five minutes of coding, ten minutes of getting up and moving around, mandatory rest.

The logic of the compatibility layer was already framed out. Under high-concurrency wake-ups, the CFS scheduler in the 2.6.32 kernel really did have a signal-loss defect. The newer kernel’s epoll and eventfd could not be used here, so he had no choice but to fall back to the lower-level combination of semaphore and spinlock. He mounted an old server image in the virtual machine and started running the stress-test scripts. The logs pouring down the terminal looked like a waterfall.

[INFO] Thread pool initialized. [INFO] Queue depth: 16. [WARN] Context switch frequency elevated.

The warning appeared right on schedule. He stared at the CPU utilization graph in the top command. The peak was stuck at 78 percent. Not enough. The production environment required it to stay stable below 65 percent. He pulled up the code and found the waiting loop around the spinlock. He lowered the busy-wait threshold from 100 microseconds to 50. Once the threshold was exceeded, it would actively call sched_yield() to give up its CPU timeslice. Recompile. Run. The curve smoothed out and dropped to 61 percent. The final line in the log popped up:

[SUCCESS] Stress test passed. 0 dropped signals.

He saved the logs and packaged them into a PDF.

Filename: v3.1_2.6.32_Compatibility_Stress_Test_Report_Lin_Chen.pdf.

By the time he finished, it was already four in the morning. In the dorm room, only his breathing and the low hum of the computer fan remained. He got up, went to the washroom, filled half a basin with cold water, soaked a towel, and laid it over his left foot. The chill temporarily suppressed the burning pain. He returned to his desk and opened the newly created folder, Spark_Technical_White_Paper.

The technical white paper required for the liaison meeting was not an academic paper. It was an engineering document. He understood that clearly. People in the industry did not care about derivations; they cared about inputs, outputs, costs, and fault tolerance. He created a new Word document.

Title: Optimization Plan for High-Concurrency I/O Queues Based on an Obsolete Kernel.

Part One: Background of the Problem. Part Two: Architecture Design. Part Three: Implementation of the Compatibility Layer. Part Four: Stress-Test Data and Resource Consumption.

He wrote quickly, without embellishment. Every paragraph corresponded to a pit he had fallen into over the past three months: rare-character encoding, SSD queue depth, kernel scheduling defects, cache-consistency latency. He turned these “errors” into “boundary conditions” in the proposal. By six in the morning, the sky outside the window had gone gray-white. He stopped and checked once through for typos and formatting errors, then exported it as a PDF.

His phone screen lit up.

Bank balance: 1835.30.

He logged into online banking and transferred 600 to the village clinic account.

Memo: Lin Xing, sodium valproate, two months.

Transaction successful. Balance: 1235.30.

Next quarter’s dorm water and electricity fee: 300. Printing costs and transportation reserve: 100. Remaining: 835.30. He ran the numbers through his head once. Everything fit together with airtight precision. He turned off the phone and leaned back in his chair. The water stain on the ceiling looked like a blurred map. He felt no excitement, only the fatigue and calm that came after finishing a step in a process.

The path from technical skill to actual money had cleared its first hurdle, but it was still far from true stability. Old Zhao’s settlement terms had not yet been finalized, and the Spark Plan liaison meeting was only an admission ticket. He had to make every step land solidly.

Friday, nine in the morning. Right on time, he sent the stress-test report to Zhao Jianguo’s internal email. Ten minutes later, the reply came:

Data reviewed. Logic is clear. Gray release into production next Wednesday; you will follow the monitoring. Liaison meeting on Saturday. Arrive at Exhibition Center Hall B one hour early for check-in.

Lin Chen replied:

Received.

He closed the laptop and began packing his bag to go out: USB drive, printed white paper, ID card, student ID, a bottle of water, two ibuprofen tablets. He put on the pair of sneakers with the unevenly worn soles. When he forced his left foot into the shoe, he still could not help sucking in a sharp breath. He adjusted the laces and shifted his center of gravity onto his right leg.

The print shop on the back street of campus had just opened. The owner, a fat man wearing reading glasses, was wiping down the counter with a rag.

“How many copies?”

“One. Double-sided. Glue binding.”

Lin Chen handed over the USB drive.

The owner plugged it into the computer and previewed the file. “Pretty clean layout. College student?”

“Mm.”

“Fifteen yuan.”

Lin Chen scanned to pay.

Balance: 1220.30.

He took the bound booklet. The cover was matte paper with a rough, grainy texture under his fingers. He opened to the first page and confirmed there was no garbled text, then turned and headed back.

On the way to the dorm, the wind was stronger than it had been the night before. The leaves on the roadside plane trees were beginning to yellow, their edges curled. He walked very slowly, avoiding cracks in the ground and puddles. His phone vibrated in his pocket. Not a text—a call.

Caller ID: Professor Zhou.

Lin Chen stopped walking. Professor Zhou had taught his data structures class when he was a sophomore. Later, he had gone to a key provincial laboratory in the capital and only occasionally returned to supervise graduation projects. He rarely called students on his own initiative.

Lin Chen answered. “Hello, Professor Zhou.”

“Lin Chen?” There was a little static in the voice on the other end, and the faint rustle of papers turning in the background. “I saw the Spark Plan list. The Science and Technology Department forwarded me the abstract of your white paper.”

Lin Chen tightened his grip on the phone. “Professor.”

“At Saturday’s liaison meeting, it won’t just be company representatives. There’ll also be two provincial industrial automation manufacturers. They won’t look at code. They’ll look at deployment cost and implementation schedule.” Professor Zhou paused, his tone even. “In your proposal, the section on hardware replacement costs is too conservative. What factories want is ‘don’t make us go through trouble,’ not ‘the optimal solution.’ Tonight, add an implementation risk assessment. Focus on downtime windows and rollback plans. Send it to me tomorrow so I can take a look. And don’t try to tough it out—an injured foot is no small matter.”

The call ended.

Lin Chen stood where he was. The wind swept through the street, stirring fallen leaves and plastic bags from the ground. He lowered his head and looked at the white paper in his hand. The title on the cover looked clear and cold in the morning light. He turned around, but not toward the dorm. Instead, he headed for the library.

Every step of his left foot came with a distinct flash of pain, like stepping on the back of a dull blade. But he knew he still had to revise another draft tonight. The road ahead was long. The only way forward was one step at a time.